A few weeks ago I went back to AI sessions I hadn’t opened for a couple of months. They held weeks of strategy and product research for my company: the reference files, the analysis, and the reasoning behind our decisions. They were gone. Deleted, with no way to get them back.
That loss hit differently than losing a file. A file you can rebuild. This you can’t. The value wasn’t only in the answers. It was in the path that led to them, and that path can never be walked again. That’s when it clicked for me. AI isn’t just another tool, like a text editor. It extends what you are capable of. The outcome matters, and so does the reasoning that got you there.
It wasn’t the first time. Earlier, my access was cut in the middle of work. The service was too busy, and users on my plan were the first to be restricted. The fix on offer was to pay more, out of nowhere. With AI you never really know how much you’ll need for a piece of work, so you can’t plan for it. I couldn’t take my work out, and I couldn’t carry on with it on my own machine. When access came back, the answers were different, and nobody told me what had changed.
What I got wrong
I made a mistake many of us make. I thought paying for AI and using it was enough. For most things it is. We lease cars, rent flats and pay for software every month, and that’s fine. But AI is not most things. It is an extension of what we can do, and that needs to be owned.
Owning something means three things:
- You can access it whenever you want.
- You can use it on your own terms.
- You can open it up and change it.
With the AI I relied on most, I had none of the three. A free tool doesn’t solve that either. Free can still be switched off, restricted or changed overnight. Only open source gives you all three.
If you can’t open it, you don’t own it.
Why it matters for all of us
AI is moving into everything we do: our work, our tasks, our processes, our lives. Something that deep inside our lives has to be under our control.
When I thought about where this is going, it scared me. Soon almost every workflow and every process will run on AI, in my company and in my personal life. I see my kids use AI every day, and mostly in good ways. That’s simply where we are. And it will shape each person’s future, personally, economically, for society as a whole, and for the environment we live in.
The economic part first. When people talk about AI replacing jobs, the answers on the table are universal basic income or a tax on robots and agents. I think both treat the symptom. They hand people a share of the value after someone else has captured it. Even if one provider made its AI free for everyone, that wouldn’t change it. That provider would still decide who uses it, how, when and on what terms. The root cause is who owns the technology. If every person owns their AI and can run it at cost, on hardware most people already have or will have anyway, paid for once, the benefits go straight to them. That changes the economic system in a way that lets everyone embrace this shift, not just the few who own the AI.
Then the planet. The IEA expects data centres to more than double their electricity use by 2030. And it isn’t only electricity. Every new data centre needs land and water, produces emissions, and comes with costs that society, the economy and taxpayers end up carrying. That isn’t good or bad in itself. It’s a consequence, and we should decide deliberately whether we want it. To decide, we first need to understand what we’re dealing with. Meanwhile, much of the compute we need already sits on our desks. What’s missing is the technology to run large models well on it, and that is a problem worth solving.
And then our independence. Today a few players own the AI that everyone increasingly depends on. That gives them power over everyone else, over what we can access, on what terms, and what the AI tells us. With something as powerful as AI, that is dangerous. Regulation can help at the edges, but it won’t fix it. Only ownership does. If everyone owns their own AI, nobody holds that power over the rest.
The good news is that owning your AI is now realistic. For years the big labs had the engines and we didn’t. According to Epoch AI, open models now trail the best closed ones by about four months, and models small enough for a single consumer graphics card catch up within a year. I believe the gap will close faster still. Models are starting to improve themselves and to train smaller versions for one person or one job. A small AI trained on what you know, running on your own machine, will serve you better than a giant one that knows everything except you.
What we’re building
In 2016 I founded City.AI to push for the responsible development and application of artificial intelligence. Ownership is the logical next step. I love technology, but I’m a people person. To me, technology only matters in how people can use it and how our environment can benefit from it. That’s why owning AI matters so much for a prosperous future for all of us.
What I want to see, and what I believe OneHill can get us to, is a world where everyone has their own AI and fully decides how it interacts with the world. Which people and organisations it talks to, which other AIs it deals with, and on what terms. Nobody can switch it off. Nobody can restrict it or change it behind your back.
Ownership is only half of it. Most AI today is built in a few places and trained largely on English. Most of the world doesn’t live there. I want AI that speaks our languages, understands our cultural backgrounds and learns from our own data, so it actually serves the people around it.
That’s why OneHill is a movement of local chapters. Each chapter pushes for ownership of AI in its city or region, and helps build AI that fits it: trained on local languages, local knowledge and local datasets, and shared openly so others can build on it. Anthill is what makes this possible. Any chapter, community or organisation can take it, train it on what matters to them, and own the result.
We made OneHill a foundation so everyone can count on it. The core stays free and open forever, with no backdoor. We sell services to keep it running, and anyone can help build it. The first piece is Anthill, an app that runs your AI on your own machine and keeps what you teach it. Along the way we built the frameworks that let us build Anthill agentically while keeping the knowledge people hold. They also help us understand what owning AI actually means in 2026.
None of this is a one-person effort. Raúl, Wilder, Cesar, William, Peter and Christophe have been building OneHill with me from the start, across engineering, research, business and community. Thank you. And thank you to everyone who has already tested Anthill, contributed to it and pushed back on our thinking.
I’d love for you to join us and help make owning AI the norm. Start by learning about Anthill, or by owning your AI today. If you want to localise OneHill, start a chapter. And if you have ideas on how we can push this forward together, get in touch.
We can argue about the good and the bad of AI, and which direction it should take. As long as one thing is clear: from today on, everybody owns their AI.
Christoph Welsbach, founding director of OneHill. 9 October 2026.
Written by Christoph. His own AI was used to help edit and structure this piece.